Skip to main content


“Don’t Override Screen Reader Pronunciation”
adrianroselli.com/2023/04/dont…

#a11y #accessibility

in reply to Adrian Roselli

I generally agree with you, however there are some rather nuanced situations where modifying how a screen reader pronounces text is quite desirable IMO. For example, MuseScore 4 deliberately modifies pronunciations of such things as Roman numeral analysis and does so with good reason.

Thread incoming.

in reply to Noah Carver Has Moved

The capitalization of Roman numerals in Roman numeral analysis is critical information for theoretical score analysis as it shows a chord's quality, thus MS4 announces when a numeral is lowercase rather than uppercase. Choosing not to modify pronunciations leaves a user unaware of this information unless they check each numeral individually, requiring them to select the numeral and enter editing mode to read character by character.
in reply to Noah Carver Has Moved

This both wastes a user's time and provides the perfect opportunity for a user to inadvertently edit what may well have been perfect analysis.
in reply to Noah Carver Has Moved

I didn't realize this was a problem until I began using MuseScore after using another notation program, Sibelius, for a semester. During my time with Sibelius, I had to constantly check Roman numeral analysis the hard way to ensure all numerals were capitalized correctly. MuseScore's announcement of lowercase Roman numerals means I can review a score quickly and efficiently in a way that wouldn't be possible if designers hadn't elected to purposefully modify pronunciations.
in reply to Noah Carver Has Moved

@noahcarver That sounds like a clear use case.

While it is less about how screen readers pronounce words and more about how they represent notation, I am also not surprised given poor mark-up support for all forms of notation (MathML being an obvious example) and screen readers lacking sufficient context for parsing.

in reply to Adrian Roselli

@noahcarver There's also a subtle but critical difference between modifying pronunciation and altering the text that is exposed. The latter should not be done to achieve the former, as it breaks braille and review by character. Unfortunately, current a11y APIs don't provide a way to distinguish these two things, though there's been more discussion about it lately in the W3C Pronunciation Task Force.
in reply to Jamie Teh

@jcsteh
Ah poop, I forgot to link that TF!

This is what I get for not sitting on a post for a few days before publishing.

@noahcarver